Breath-Controlled Wind Synth Expressivity

ISEF Category: Technology Enhances the Arts

Subcategory: Music and Image Manipulation  ·  Difficulty: Advanced  ·  Setup: School Lab  ·  Time: 1 to 2 Months

The Hook

A tiny change in your breath can change a phrase more than a new note can. That is why wind players sound human, not robotic. You can turn that idea into a science fair project. The trick is to measure breath like a signal, not just a puff of air.

What Is It?

This project studies how a breath-controlled music interface can capture the details that make playing feel expressive. In a wind instrument, you do more than blow harder or softer. You also add noise, shape the attack, and vary pressure in ways that change tone and feel. Your device tries to sense those cues with a mic inside a mouthpiece, then turns them into MIDI control changes, which are the messages that tell digital music software how to behave.

Think of it like translating from one language to another. Your breath is one language, and MIDI is the other. A simple controller might only hear loud or soft. Your system tries to hear more, like turbulence, onset shape, and pressure pattern, then map those features to musical expression. You can compare your design against a commercial breath controller by asking musicians which one feels more natural, responsive, and musical.

Why This Is a Good Topic

This is a strong science fair topic because you can test a clear input, a clear output, and a real user experience. You can change the sensing method, the signal features, or the mapping model, then measure how those changes affect expressivity. The project connects to assistive music tech, digital instruments, and human-computer interaction, so it has a real audience. You can also learn signal processing, model training, and experimental design without needing a university lab.

Research Questions

• How does using both breath pressure and turbulence noise affect musician-rated expressivity compared with using pressure alone?
• What is the effect of different MIDI CC mapping strategies on perceived responsiveness during performance?
• Does a small MLP produce more natural control than a linear mapping for breath to MIDI output?
• To what extent does mouthpiece geometry change the stability of the sensor signal during repeated performances?
• Which breath features best predict musician ratings of expressivity, timing feel, and control?
• How does the prototype compare with a commercial breath controller on expressivity ratings and consistency?

Basic Materials

• INMP441 microphone module or similar I2S digital microphone.
• ESP32 microcontroller or similar board with I2S support.
• 3D-printed mouthpiece or prototype housing.
• USB cable for programming and data transfer.
• Laptop with a DAW or MIDI monitoring software.
• Breadboard and jumper wires.
• Small set of resistors and connectors as needed for the build.
• Headphones or studio monitor speakers for listening tests.
• Smartphone or camera for documenting trials.
• Consent forms and rating sheets for musician feedback.

Advanced Materials

• High-quality I2S microphone modules for side-by-side sensor comparison.
• Commercial breath controller for baseline testing.
• 3D printer with multiple mouthpiece designs.
• Audio interface for low-latency monitoring.
• Calibration pump or pressure reference setup.
• Reference pressure sensor for ground-truth comparison.
• Laptop with Python, NumPy, pandas, scikit-learn, and Matplotlib.
• MIDI monitor software for logging control data.
• DAW with MIDI learn support.
• Statistical analysis software for rating comparisons and inter-rater agreement.

Software & Tools

• Python: Cleans sensor data, extracts breath features, and trains the mapping model.
• Jupyter Notebook: Lets you document experiments, plots, and analysis in one place.
• ImageJ: Measures 3D-printed mouthpiece dimensions if you compare prototype shapes.
• MIDI-OX: Monitors MIDI CC output and checks whether the controller behaves as expected.
• Audacity: Reviews recorded performance clips and helps you compare signal quality across trials.

Experiment Steps

1. Define the exact musical behaviors you want to measure, such as responsiveness, smoothness, and expressive range.
2. Choose the breath features you will extract first, then decide how each feature should control one or more MIDI parameters.
3. Build a baseline version with a simple mapping so you can compare it against your machine learning version.
4. Plan a fair comparison against the commercial breath controller, including the same musical task and the same scoring rubric.
5. Design your data collection so you can separate sensor quality, mapping quality, and player preference.
6. Set up your analysis before testing, so you know which ratings, signals, and summary metrics will answer your research question.

Common Pitfalls

• Using an unsealed mouthpiece, which lets breath leak and makes the signal jump around.
• Training the MLP on too few performances, which makes the model memorize one player instead of general breath patterns.
• Comparing controllers with different MIDI settings, which makes the expressivity test unfair.
• Rating expressivity without blind listening, which lets players guess which device they used and bias the scores.
• Ignoring latency between breath input and MIDI output, which can make a technically accurate controller feel awkward to play.

What Makes This Competitive

A strong version of this project does more than build a working controller. It tests whether a richer breath signal really improves musical control, and it proves that with careful comparisons. You can get stronger results by using blind musician ratings, multiple control baselines, and stats that measure both accuracy and user preference. A project like this stands out when the analysis connects signal features to how musicians actually experience expression.

Project Variations

• Compare different mouthpiece shapes to see which one captures breath turbulence most cleanly.
• Test whether one sensor type, such as a mic versus a pressure sensor, gives better expressive control.
• Train separate mappings for different musical tasks, such as sustained notes versus fast phrasing, and compare which feels more playable.

Learn More

• NIH PubMed: Search for review articles on breath sensing, digital musical instruments, and human-computer interaction in music.
• IEEE Xplore: Search for papers on MIDI control, expressive performance interfaces, and musical instrument design.
• MIT OpenCourseWare: Look for free courses on signal processing, machine learning, and digital signal systems.
• NASA NTRS: Search for signal processing and sensor fusion papers that can give you ideas for feature extraction and noise handling.
• Python documentation: Read the NumPy, pandas, and scikit-learn guides for basic data cleaning, modeling, and evaluation.
• The Computer Music Journal: Search the journal for research on expressive controllers and performance interfaces.